3D Gaussian Splatting! - Computerphile

Channel:Computerphile · Length:17:39 · Published:2024-03-14 · Views:276743

Watch on YouTube

Summary

A Computerphile video in which Dr Mike Pound and PhD student Lewis Stuart explain 3D Gaussian splatting, a quick technique that turns photos of a scene into a 3D model without much compute. It includes a demo and compares the approach to NeRFs.

Why HN recommends it

HN recommends it as an accessible introduction to Gaussian splats, made by people who understand the subject.

Here's a much better explainer from some people who seem to actually understand the subject matter and not just the buzzwords — HN, 2024

Computer vision, speech & NLP · Computer graphics & rendering

Subtitles (excerpt)

um now you may look at that and go that's all right it's not great but what happens when I increase the quality boom now that to me looks very very nice and the most impressive part is if we move it around look how quick it is you know granted my computer is not the best but you could not do this with a Nerf in the previous video whenever I moved the camera to a new position I'd have to wait about 5 seconds for it to render a proper image this is you know pretty much instant granted as I say I haven't got the best computer but if you had the latest and the best you could get 100 FPS on this and you're also running off a web server right which is going to me way slower sending images retrieving images yeah let's have a very quick look at Nerf just in one minute from the previous video and then we can talk about something that's totally different which is called gussian splatting what they do is they take a series of RGB images and from those RGB images of a scene they're able to reconstruct it in 3D using a neural network so if you remember in Nerf what we have is we have a scene that's in 3D and maybe we have an object that we're trying to reconstruct in our scenes we did a Christmas tree in our video so

On Hacker News

You're on HN, not reddit. OP linked to a special interest blog, not an explainer piece. The burden to catch up here is on the reader. The search input is right at the bottom: [LINK] but start here: TUM AI Lecture Series - The 3D Gaussian Splatting Adventure: Past, Present, Futur (George Drettakis) [LINK] or here: [VIDEO] — mxfh, 2026-01
Not a reading guide, but Computerphile have a good introductory video on Gaussian Splats: [VIDEO] Individualkex also has a couple videos on the high level ideas: [LINK] — yokto, 2024-11
Here's a much better explainer from some people who seem to actually understand the subject matter and not just the buzzwords (and yes, the end result is basically a highly-detailed point cloud): [VIDEO] — ghostly_s, 2024-05

Related videos